This application relates to apparatus and methods for automatically determining and providing recommendations of items to advertise customers. In some examples, a computing device generates feature data based on historical website interaction data, historical transaction data, and item categorical data. The computing device trains each of a plurality of machine learning models based on the generated feature data. The computing device may then receive a plurality of recommended items to advertise in association with an anchor item. The computing device may execute the trained machine learning process to generate prediction data associated with a future time period. The prediction data may identify a number of times each recommended item may be purchased during the future time period. The computing device may then rank the plurality of recommended items based on the prediction data. In some examples, the computing device filters the plurality of recommended items based on item categories.
Legal claims defining the scope of protection, as filed with the USPTO.
2. The system of claim 1, wherein the feature data includes interaction features for each recommended item of the plurality of recommended items based on the session activity data.
4. The system of claim 3, wherein the feature data includes categorical item features for each recommended item of the plurality of recommended items, the categorical item feature comprising a taxonomical hierarchy.
5. The system of claim 1, wherein the prediction data for each of the recommended items identifies a predicted number of times each recommended item will be purchased during a future time period.
6. The system of claim 1, wherein generating the feature data based on the session activity data comprises generating a feature vector for each recommended item of the plurality of recommended items, wherein the feature vector for each recommended item identifies the anchor item, the recommended item, and the features.
7. The system of claim 6, wherein the features include aggregated co-count, contextual, categorical, and aggregated view-buy count features.
8. The system of claim 1, wherein filtering the ranking of the plurality of recommended items further comprises: removing a recommended item of the plurality of recommended items from the ranking of the plurality of recommended items if at least one corresponding distance value is not at least a threshold value.
9. The system of claim 1, wherein the computing device is configured to train the trained machine learning process with interaction data, popularity data, and categorical data for a first period of time.
10. The system of claim 9, wherein the computing device is configured to periodically train the trained machine learning process with the interaction data, the popularity data, and the categorical data for a second period of time, wherein the second period of time is less than the first period of time.
11. The system of claim 1, wherein the trained machine learning process comprises a plurality of machine learning models, wherein each machine learning model is trained with updated feature data of a different determined item type.
13. The method of claim 12, wherein the prediction data for each of the recommended items identifies a predicted number of times each recommended item will be purchased during a future time period.
14. The method of claim 12, wherein the categorical attribute embedding uses the session activity data to determine item categories, and wherein filtering the ranking of the plurality of recommended items is based on the categorical attribute embeddings.
15. The method of claim 12, further comprising training the trained machine learning process with interaction data, popularity data, and categorical data, wherein popularity data includes at least one of ratings and reviews of each recommended item.
17. The non-transitory computer readable medium of claim 16, wherein the prediction data for each of the recommended items identifies a predicted number of times each recommended item will be purchased during a future time period.
18. The non-transitory computer readable medium of claim 16, wherein the operations comprise training the trained machine learning process with interaction data, popularity data, and categorical data.
Cooperative Patent Classification codes for this invention. Click any code to explore related patents in that topic.
December 20, 2019
September 27, 2022
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.